Prompting
System Prompt
The high-level instructions given to an AI model at the start of a conversation that define its role, behaviour, and constraints.
01 ——
In plain English
A system prompt is a special set of instructions given to an AI model before any user input. It tells the model who it is, what it should do, what it should refuse, and how to respond. Users typically don't see it.
Common system-prompt elements:
- Role — "You are a friendly customer support agent for Acme Co."
- Tone — "Reply in a warm, professional tone."
- Constraints — "Never discuss competitor products."
- Format — "Always respond in Markdown with clear headers."
- Tools — describing what functions the model can call
Why it matters: System prompts are the main way developers customise general-purpose models for specific products. A well-written system prompt can transform GPT-4 into a coding tutor, a legal assistant, a sales coach, or a children's storyteller — without any retraining.
System prompts are also the first place to look when an AI tool misbehaves.
02 ——
Related terms
Prompt Engineering
The practice of crafting inputs to an AI model carefully to get better, more reliable outputs.
LLM
Large Language Model — the type of AI behind tools like ChatGPT and Claude, trained to understand and generate text.
Guardrails
Rules and filters that constrain what an AI model can output — used to block harmful, off-topic, or non-compliant responses.
Jailbreak
A prompt or technique that tricks an AI model into ignoring its safety rules and producing content it would normally refuse.
Chain of Thought
A prompting technique where you ask the AI to "think step by step" before giving an answer — usually leading to better reasoning.
Back to glossaryLast reviewed May 2026